Drazin inversion in the von Neumann algebra generated by two orthogonal projections (Q2389267)

The authors establish criteria for Drazin and Moore-Penrose invertibility of operators in the von Neumann algebra generated by two orthogonal projections and present explicit representations for the corresponding generalized inverses. They also illustrate the results by means of several examples considered recently in the literature. The details are as follows. Let \(\mathcal H\) be a Hilbert space and \(\mathcal B (\mathcal H)\) be the algebra of all bounded linear operators acting on \(\mathcal H\). For \(A \in \mathcal B ({\mathcal H})\), a Moore-Penrose inverse is an operator \(X \in \mathcal B ({\mathcal H})\) such that \(AXA = A\), \(XAX = X\), \((AX)^* =AX\), \((XA)^* = XA\). Such an \(X\) exists if and only if the range of \(A\) is closed, in which case it is unique. The Moore-Penrose inverse of \(A\) is denoted by \(A^{\dagger}\). Let \(k\) be the index of \(A\). Then the Drazin inverse of \(A\), denoted by \(A^D\), is the unique operator \(X \in \mathcal B ({\mathcal H})\) that satisfies \(A^{k+1}X = A^k\), \(XAX = X\), \(AX=XA\). These are particular instances of what are called generalized inverses. \(A\) is said to be Moore-Penrose invertible or Drazin invertible, if the Moore-Penrose inverse or the Drazin inverse of \(A\), respectively, exists. Again, let \(\mathcal H\) be a Hilbert space, \(P,Q \in \mathcal B (\mathcal H)\) be orthogonal projections, and let \(W^*(P,Q)\) denote the smallest von~Neumann subalagebra of \(\mathcal B (\mathcal H)\) that contains \(P,Q\) and the identity operator \(I\). Let \(L\) and \(N\) denote the range spaces of \(P\) and \(Q\), respectively. For a closed subspace \(M\) of \(\mathcal H\), let \(P_M\) denote the orthogonal projection of \(\mathcal H\) onto \(M\). Define the subspaces \(M_{11}=L \cap N\), \(M_{10}=L \cap N^{\perp}\), \(M_{01}=L^{\perp} \cap M,\) and \(M_{00}=L^{\perp} \cap N^{\perp}\). Let \(H\) denote the compression of \(I-P_N\) to the subspace \(M_0=L \ominus (M_{11} \oplus M_{10})\) and \(\varphi_{ij}\) be Borel-measurable and essentially bunded functions on the spectrum \(\sigma(H)\) of \(H\). Set \(\Phi_A= \begin{pmatrix} \varphi_{00} & \varphi_{01}\\ \varphi_{10} & \varphi_{11} \end{pmatrix}, ~\omega_A = \varphi_{00}\varphi_{11} - \varphi_{01}\varphi_{10}\) and \(\varphi_A ={\mid \varphi_{00} \mid }^2 + {\mid \varphi_{01} \mid}^2 + {\mid \varphi_{10} \mid}^2 + {\mid \varphi_{11} \mid}^2\). Denote by \(\bigtriangleup_r(A)\) the set of all \(t \in \sigma(H)\) such that the rank of \(\Phi_A(t)\) equals \(r (r=0,1,2)\). Set \(\bigtriangleup_{10}(A)=\{t \in \bigtriangleup_1(A): \text{tr}(\Phi_A(t))=0\}\) and \(\bigtriangleup_{11}(A)= \bigtriangleup_1(A) \backslash\bigtriangleup_{10}(A)\). Let \(A \in W^*(P,Q)\). The authors first prove that the range of \(A\) is closed if and only if \(\omega_A\) and \(\varphi_A\) are separated from zero on \(\bigtriangleup_2(A)\) and \(\bigtriangleup_1(A)\), respectively. Next, it is shown that \(A\) is Drazin invertible if and only if \(\omega_A\) and \(\text{tr} \varphi_A\) are separated from zero on \(\bigtriangleup_2(A)\) and \(\bigtriangleup_{11}(A)\), respectively. A simplification is presented next, when \(A\) is a polynomial in \(P\) and \(Q\). The authors then proceed to apply their results to show their applicability in concrete examples, studied recently in the literature. We present a sample result, in this direction. The following are equivalent: (i) \(P+Q\) is Drazin invertible, (ii) \(P+Q\) is Moore-Penrose invertible, (iii) \(M_0=\{0\}\) or \(M_0 \neq \{0\}\) and \(H\) is invertible. Explicit representations of the generalized inverses concerned are also provided.